• 作者： Too, J.R.; Chen, Biau-Dar
• 中文摘要： 本研究對線上學習神經控制器做探討。當網路拓蹼被選定後,所需調整的參數僅有一個,即網路的學習速率,此神經網路控制器的學習樣本少,僅需初始一組啟動的樣本,而且可以線上學習控制。網路的訓練是採用移動視窗學習法則,將神經網路控制器所決定的控制值改變量,納入下一次神經網路控制器學習的樣本,以做為適應控制的準備。此一線上學習神經網路控制器已用來模擬控制處理廢汙水廠的活性汙泥曝氣槽中迴流汙泥之迴流比。線上學習控制器對於不穩定的非線性系統,有著適應控制的特性。
• 英文摘要： In this study, an on-line training neural controller is investigated in this study. Once the topology of the neural network has been selected, one parameter, only the learning rate is left to be tuned . This neural controller only needs a start-up datum to train the network. The neural network controller may be trained on-line with a moving- window learning rule. In this mode of training, the previous output from the neural controller is used as a training datum for the next run. The on-line learning neural controller has been applied to regulate the reflux ratio of an activated sludge aerator in a wastewater treatment plant. The proposed control mode can handle both servo and regulator problems. The on-line learning controller also possesses the characteristics of an adaptive control for an unsteady and non-linear system.
• 中文關鍵字： 類神經控制器; 程序; 控制; 汙泥; 模擬
• 英文關鍵字： Neural Controller; Process; Control; Sludge; Simulation